AI-Assisted Causal Pathway Diagram for Human-Centered Design

Document Title

AI-Assisted Causal Pathway Diagram for Human-Centered Design

Document Information

• Subject Area: Human-Computer Interaction (HCI), Design Support Tools
• Keywords: Causal Pathway Diagram, Human-Centered Design, Generative AI, Large Language Models (LLM), Implementation Science

Research Background and Problem

• Problem or Challenge:

  • Human-Centered Design (HCD) emphasizes solving design problems with a user-centered approach, but there is currently a lack of tools that systematically integrate theory with design practice.
  • While Causal Pathway Diagrams (CPD) have been successfully applied in implementation science, their complexity and implementation costs may limit their use in design practice.
  • In addition to barriers in applying design theory, designers often struggle to visually represent user research findings due to knowledge domain limitations.

• Significance:

  • Efficient theory-driven design helps align user goals with design outcomes, reduces the risk of design errors, and enhances design effectiveness and user satisfaction.
  • By introducing artificial intelligence, particularly Generative AI, operational complexity can be reduced, fostering creativity and optimizing decision-making for designers.

• Research Motivation and Related Work:

  • Implementation science provides the theoretical foundation for CPD, a graphical tool that helps designers establish causal relationships between user needs and design outcomes.
  • Developing an AI-supported plugin could address usage barriers, significantly reducing the cognitive load on designers when learning and applying CPD, and improving design efficiency.

Solution

• Method or Solution:

  • Designed and developed a CPD plugin tool integrated into the online collaboration platform Miro, incorporating support from Generative AI (GPT-4).
  • The plugin uses a predefined CPD syntax framework and chart component tools to help designers easily create and iterate CPDs while providing real-time suggestions.

• Innovative Features:

  • Provides AI-driven real-time guidance for CPD graphical tools, enabling users to receive theoretical and creative support during the design process.
  • The plugin not only reduces the technical barriers to learning CPD creation but also introduces features to enhance design creativity and provide evidence-based support.

• Implementation Steps:

  1. Component Generation: Users can drag and drop different types of CPD components (e.g., strategies, evaluation mechanisms, barriers, goals), with the plugin automatically defining symbols for each component and supporting users in explaining their purpose.
  2. Guided System: The plugin offers a step-by-step guide to help designers create a complete CPD, starting from the end goal and working backward, generating recommendations at each step.
  3. Creative Expansion: Through AI-generated suggestions, designers can expand on existing components with different design strategies or pathways.
  4. Pathway Validation: The plugin can verify whether the user-generated CPD is syntactically correct, such as ensuring component connections are complete and sequences are logical.
  5. Component Explanation: Users can quickly access definitions and additional background information for components, reducing the learning curve.

Research Outcomes

• Specific Outcomes:

  • Theoretical Design Support: CPD significantly aids designers in conducting systematic and goal-oriented design during the early stages.
  • Creative Generation: Designers using the plugin can generate more and richer design pathways compared to traditional methods, significantly enhancing creativity.
  • Reduced Cognitive Load: The CPD plugin reduces the burden on designers to memorize theoretical frameworks and manually create content through automation tools and real-time AI suggestions.
  • Structured Communication: CPD facilitates easier communication of ideas between design teams and business decision-makers, improving collaboration efficiency through a unified causal diagram language.

• Advantages and Comparisons:

  • Compared to non-plugin methods, the plugin accelerates the design process and increases user confidence in the accuracy and practicality of design outcomes.
  • While using the plugin slightly increases the time required to generate CPDs, the number of pathways generated increases significantly.

• Experimental or Evaluation Results:

  • Experiments showed significant improvements in design experience and output quality when designers used the plugin (e.g., pathway generation increased by ~58%).
  • The plugin effectively reduced operational difficulty and was deemed highly effective in fostering creativity and structuring design agendas.

• Limitations and Future Directions:

  • Limitations:
    • AI-generated recommendations may appear overly generic due to insufficient contextual information, requiring further model improvements for more specific suggestions.
    • The lack of transparency in the information sources for AI suggestions raises concerns about the credibility of recommendations, which needs to be addressed.
  • Future Directions:
    • Build a richer causal pathway database to generate more reliable, evidence-supported suggestions.
    • Explore seamless integration of the plugin into actual design workflows, including later design stages.
    • Expand the plugin's application to support in-person design meetings, enhancing team interaction and formalizing creative expression.

This study proposes an innovative approach to introducing CPD into Human-Centered Design and simplifying its application through Generative AI, laying the foundation for a deeper integration of design methodology and technology.